ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

[pre-commit.ci] auto fixes from pre-commit.com hooks

Browse source 
for more information, see https://pre-commit.ci
This commit is contained in:
pre-commit-ci[bot] 2026-03-14 08:28:31 +00:00 • committed by Daniel Han
commit 7202f81985
1 changed files with 14 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

21
studio/backend/core/inference/llama_cpp.py
View file

@ -344,13 +344,20 @@ class LlamaCppBackend:
cmd = [
binary,
"-m", model_path,
"--port", str(self._port),
"-c", str(n_ctx),
"-ngl", str(n_gpu_layers),
"--parallel", "1", # Single-user studio, saves VRAM
"--flash-attn", "on", # Force flash attention for speed
"--fit", "on", # Auto-fit to available device memory
"-m",
model_path,
"--port",
str(self._port),
"-c",
str(n_ctx),
"-ngl",
str(n_gpu_layers),
"--parallel",
"1", # Single-user studio, saves VRAM
"--flash-attn",
"on", # Force flash attention for speed
"--fit",
"on", # Auto-fit to available device memory
]
if n_threads is not None: